Machine learning-based human-robot interaction in ITS

(1)

Information Processing and Management 59 (2022) 102750

Available online 16 October 2021

Machine learning-based human-robot interaction in ITS

Jingyao Wang

^a^,^*

, Manas Ranjan Pradhan

^b

, Nallappan Gunasekaran

^c

aZhejiang Normal University, laboratory construction and Equipment Management Office, 321004 Jinhua, Zhejiang, China

bSchool of IT, Skyline University College, Sharjah, UAE

cComputational Intelligence Laboratory, Toyoto Technological Institute, Nagoya, Japan

A R T I C L E I N F O Keywords:

Human-computer interaction Machine learning

Intelligent transportation system Intelligent traffic monitoring system

A B S T R A C T

In the last few years, intelligent transport systems (ITS) have drawn growing attention, and these applications would have a clear and more comfortable experience for transportation. ITS provides applications with a chance to address the future condition on the route beforehand. The major issues in ITS to accomplish a precise and effective traffic flow prediction system are essential.

Therefore, in this paper, a machine learning-assisted intelligent traffic monitoring system (ML- ITMS) has proposed improving transportation protection and reliability to tackle several challenges. The suggested ML-ITMS uses mathematical models to improve the accuracy estimation of traffic flow and nonparametric processes. The Machine Learning-based (ML) method is one of the best-known methods of nonparametric. It requires less prior information about connections between various traffic patterns, minor estimation limitations, and better suitability of nonlinear traffic data features. Human-Robot Interaction (HRI) helps resolve crucial issues concurrently on both the customers and service supplier levels at both ends of the transport system. Thus the experimental results show the proposed ML-ITMS to enhance traffic monitoring to 98.6% and better traffic flow prediction systems than other existing methods.

1. Introduction to human-robot interaction in ITS

Nowadays, innovation can appear to do surgical operations to domestic cleanings, and robots are currently doing several tasks (Shahriar et al., 2018, June). Encountered with such amazing advances, people prefer to ignore the innovation that still has its limitations (Manogaran, Shakeel, Priyan, Chilamkurti, & Srivastava, 2019). In social robots, people can overlook the real robot capacities and trustworthiness, and even clear proof of robot boundaries is not a successful preventive measure (Nguyen et al., 2016). By operating together, including human beings in different fields, the robots often alter the social and technical lives (Gao et al., 2020, August). Owing to the ever-increasing need for robots for human interaction, cooperation, and support, HRI presents new security, automation, and recognition problems (Jindal et al., 2019). Intelligent Transport Systems (ITS) are control and IT Systems using in- tegrated communications and data processing technologies to enhance the mobility of human beings and goods, safety increase, traffic congestion reduction, and effective management of incidents. The involvement between humans, robots, and the ecosystem should be considered robustly in medical situations (Manogaran et al., 2021). The robot needs to classify the atmosphere and the condition of the supported user before executing an operation (Priyan & Devi, 2018). Interactive robots share the same workplace and work with individual co-workers (Shakeel, Baskar, & Selvakumar, 2019). Such an interactive situation enables robots and people to utilize the

* Corresponding author.

E-mail addresses: [email protected] (J. Wang), [email protected] (M.R. Pradhan), [email protected] (N. Gunasekaran).

Contents lists available at ScienceDirect

Information Processing and Management

journal homepage: www.elsevier.com/locate/infoproman

https://doi.org/10.1016/j.ipm.2021.102750

Received 21 May 2021; Received in revised form 21 August 2021; Accepted 2 September 2021

(2)

best capabilities. The creation and recognition in the enterprise of interactive robots depend heavily on the efficient, logical communication of human robots (Ramprasad & Amudha, 2014, February). The ITS aims to achieve efficiencies by minimizing traffic problems; it aims to reduce travel time and improve safety and comfort. Traffic congestion and information are not limited and to road safety and effective use of infrastructure. Robotic systems and people need to recognize each other and communicate intuitively (Gao, Wang, & Shen, 2020). It would allow interactive robots in manufacturing industries that struggle to produce ever more flexibly due to requests from consumers for specialized goods (Sivaparthipan et al., 2019). Traffic and prediction systems can improve conditions and reduce travel delays by improving the use of available capacities. These systems use existing, and emerging technologies to monitor, manage and control the transport systems, such as computer, communication, and control. They provide system users with different levels of traffic information, including many ITS providers, so that passengers can make quick and informed decisions about travel.

The development of new robots, architectures, and interaction processes, grounded in robotics, information science, and design research, is part of the Human Robotic Interaction area (Elhoseny et al., 2017). Human-Robot Interaction covers a broad range of academic fields and academic activities (Vasconez et al., 2019). To avoid road congestion and traffic problems, the HRI model is developed (Guo et al., 2020). An effective and reliable transport system can aid in organizing transport services more effectively, scattering traffic flows until it is crowded and even making road recreation more plentiful (Fu et al., 2020). Transports are getting more difficult with developments in urban development and cars’ popularity: traffic flows are congested, regular accidents, and road conditions worsen. Interaction between human robots is a study of human and robot interactions. Researchers are often referred to as Human-Robot Interaction is a multidisciplinary field with contributions from interactions between people and computer, artificial intelligence, robotics, understanding natural languages, design, and psychology.

The most popular method to reduce the traffic congestion of such schemes is the intelligent Transportation System (ITS) (Sirohi et al., 2020, Shakeel et al., 2018). ITS is a dynamic system implemented by advanced systems such as communications networks for road transport. ITS is indeed able to boost the quality, mitigate traffic congestion expand road ability and decrease road traffic incidents and industrial pollution through the implementation of the communication network, plentiful on-road sensors, etc. (Nasimi, Habibi, & Schotten, 2020).

An effective and powerful traffic forecasting system could provide continual and exact on-road position focused on past traffic situations as an essential part of ITS (ITS, Preeth et al., 2020). This knowledge can benefit ITS implementations, including road congestion management, light traffic management, vehicle cloud, etc. (Sheikh, Liang, & Wang, 2020). One challenge in introducing and sustaining a vehicle cloud is to compute the usable rolling stock on a provided road segment to assess the cloud’s workflow. Road supplies are mostly collected from the roadway or municipal motor vehicles, making it critical for the cloud scheme to detect how often vehicles are on the provided road network in the future. In many other words, over the period, the interaction among various modes of traffic determines the amount of traffic. In particular, long-term traffic characteristics influence temporal patterns and correlations between the human-robot interaction. ITS offers the opportunity to deal in advance with the potential road situation. ML-ITMS can overcome the main problems with ITS to implement an accurate and efficient traffic flow control system. In the recommended ML-ITMS mathematical model, road traffic and non - parametric systems are estimated accurately. It needs fewer details on interactions between different traffic patterns, more periodic estimates, and better suitability for nonlinear traffic data. Therefore, the proposed ML-ITMS demonstrates experimental results to improve traffic control and better prediction mechanisms for traffic flow than other existing techniques.

The remaining article is organized as follows: Section 2 comprises various background studies concerning HRI. Section 3 Elaborates the proposed ML-ITMS model to enhance transportation security and accuracy to tackle several challenges. Section 4 constitutes the results that validate the performance with the corresponding descriptions. Finally, the conclusion with future perspectives is discussed in section 5.

2. Background study on psychological abuse and depression

This section discusses several works that various researchers have carried out; Kai Lin et al. (Lin et al., 2020) introduced a Hybrid Body Sensor network architecture based on Multi-Sensor Fusion (HBMF). HBMFpromoted the most innovative intelligent healthcare services, combining different devices, interactions, robots, and data management innovations. In particular, the multi-sensor neural network fusion process for sensor networks was analyzed for improving the efficiency of configuration decisions in the medical HRI scenario. The architecture guaranteed the efficiency and quality of the system in the clinical HRI case compared with existing multi-sensor convergence procedures.

Mahdi Khoramshahi et al. (Khoramshahi & Billard, 2019) developed Task-Adaptation in Physical Human-Robot Interaction (TA-HRI). TA-HRI allowed robots for the physical human-robot interaction, constructively and compliantly adjusting a human person’s movements. TA-HRI used a class of parameterized functional systems which made the transition among encrypted tasks quick and scalable. An exhaustive analysis of the TA-HRI approach was included concerning stability, consistency, and optimity to provide a secure and logical human interaction behavior.

Pedro Neto et al. (Neto et al., 2019) proposed Gesture-based human-robot interaction (GB-HRI). GB-HRI could help a human employee to transfer materials and equipment and keep objects for assembly operations. Static and dynamic data blocks were provided a static, dynamic, composite movement classification to the artificial neural network (ANN). GB-HRI suggested an automatic task manager parameterization for the HRI application, wherein the co-worker chose/strengthened robotic choices by gestures as per machine voice and visual feedback. Observations in an assembly process showed the performance of the suggested solution.

Lorenzo Desideri et al. (Desideri et al., 2019) discussed Emotional Processes in Human-Robot Interaction (EP-HCI). EP-HCI aimed to establish whether an interplay-related emotional process was created through a brief screening evaluation carried out by a robot

(3)

compared to a skillful physician. Likewise, mental and task output did not change under different conditions. Even then, EP-HCI examination of non-verbal actions found that respondents spent further hours staring at the robot and did less glance aversion than the human investigator when interacted with the robot.

Juan Pablo Vasconez et al. (Juan Pablo Vasconez et al) introduced the Human-Robot Interaction strategy for Commercial Vehicle Driving (HRI- CVD). HRI-CVD focused on creating a human-robot framework for Renault Twizy, which could utilize behavioral criteria to enhance driver safety throughout night and day activities. Based on the mental state variables, a human-robot interaction scheme was proposed that limits the velocity and breaking of a Renault Twizy vehicle. The proposed strategy of human-robot interaction could improve security for users and tourists during drive activities.

Daniel Ullrich et al. (Ullrich et al., 2021) proposed an Empirical Simulation and Psychological Analysis (ES-PA) to overcome the serious issue, particularly when it comes to personal well-being. Therefore, insights into the production and influence of over- confidence factors in robotics were a significant factor in defensive measures by the authors. ES-PA incorporated such results into a conceptual model over time, connecting with associated psychological principles such as positivism, immediate remuneration, improper widespread, and human social principles. The present ES-PA discussed the limitations and consequences of robot design and future studies. Based on the survey, to overcome all the ITS problems and achieve an accurate and efficient traffic prediction method.

ML-ITMS has been proposed for enhancing transport safety and efficiency by a Machine Learning method to overcome all the challenges.

3. Machine learning-assisted intelligent traffic monitoring system (ML-ITMS)

This section discusses the intelligent transportation system using machine learning for traffic flow prediction. Intelligent transportation systems (ITS) have attracted increasing interest, with clear and more convenient travel experiences in these implementations. ITS allows applications to deal with the potential conditions in advance on the pathway. The main questions in ITS are important to achieve a detailed and efficient method for traffic control. Therefore, ML-ITMS has been proposed to evaluate the traffic congestion in the roadside units to improve ITS performance. This paper provides a short-term ML model of traffic flow and optimizes Support vector machine (SVM) parameters to enhance traffic flow prediction. The goal monitoring system introduced in this paper has considerably increased the precision of the count. The proposed ML-ITMS, SVM, and RF, are specially built for LoRa in one inves- tigation. Intelligent sensors are used for data collection and are then migrated to the LoRa cloud server. Lora (Long-Range Wide Area Network) is another common approach we found during this survey is a LIDAR technology (Light Detection and Ranging). A data processing method is then used as feedback for ML-ITMS. The platform then passes by ML-ITMS functions include traffic prediction by LIDAR, public security safety of cities, provision of medical facilities, parking control. Dynamic flow and static traffic signals are the two main issues behind traffic congestion. This document, therefore, aims at a road preview by LoRa, known as Long Range Wide Area Network Technology, to resolve this problem. LIDAR offers very low bit rate long-range connectivity. This technology provides an intelligent system with good traffic control ideal for efficient management and time usage. In recent studies, several ideas from wireless networking, traffic philosophy, and machine learning have been incorporated.

Some collaborative robots can even be trained to do tasks in logistic applications by letting other people guide their weapons once to get the move. This reduces inadequate programming time and speeds of the personalized packaging process. Robots are entering the logistics and transport sectors rapidly.

Figure 1 elaborates the traffic management system. This figure shows the general functioning of the proposed scheme. The information obtained from intelligent sensors is migrated to the LoRa cloud platform. The platform then interprets the data and the algorithm for machine learning, used as the ITMS input. Until a solid ITMS architecture is constructed, ITMS can incorporate

Fig. 1.Traffic management system.

(4)

applications like adaptive navigation systems. This research is, therefore, ultimately aimed at laying the foundation for ITMS.ITMS includes traffic flow management through installed signals from LoRa, vehicle parking, health care, and city protection. It is very important to collect the data, properly handle it, and predict the best possible response. The two traffic signals located at each street corner transfer data to the cloud server. The data are submitted to the learning algorithm of the computer, and it is forecast whether or not the path is congested. It is one of the algorithms for master learning most administered and controlled. Smart traffic control has many valuable benefits, such as reducing emissions and fuel consumption. The device can assist in the event of an accident and provide the driver with the shortest route. Robots provide advantages in many areas, including logistics and transport safety and efficacy. The use of robots will also help warehousing employees in the industry and end customers since the service is much faster and more straightforward.

Figure 2 explores the machine learning-based traffic congestion monitoring scheme. The layout of autonomous vehicles and how autonomous cars cope with congestion are shown in the above figure. Data from different sources are gathered here. In the meantime, drivers who are about to start or expect to hit the traffic are involved in learning the existing traffic scenarios to determine whether to get on with the road and which way to save time and congestion is the shortest path possible. Distribution networks throughout the global supply chain require a high number of different and complex tasks. This presents automation challenges that are easier and cheaper to carry out where repetitive tasks are easier. However, new technologies in the logistics field overcome these barriers in several ways. The streaming of real-time data is the process in which large amounts of data are processed fast, such that a company that extracts information from these data can respond in real-time to changing conditions. The intelligent transport system will collect a variety of roads. Details from the number and average velocities of vehicles that reach a certain point. In the case of ITS technology, vehicles can even be positioned via mobile telecommunications or satellite systems. A base station is a radio receiver/transmitter in wireless computer networking that acts as the backbone for the local wireless network and the portal between a wired network and the wireless network. The transmitter and wireless router usually are low-power. The intelligence systems are state-of-the-art technologies aimed at providing creative services for multiple means of transport and traffic management modes and making transport networks smoother, more coordinated, and smarter for different users. Transport service provider means the supply of applications or connected facilities allow occupants to access and use apps and related services for and inside a vehicle or vehicles used for the transportation and transport of occupant(s). Transport service provider means any person whose main business is to provide transport to or for passengers in vehicles.

Figure 3 shows the smart transport system’s layered architecture. The bottom layer comprises data processing components that include various sensors (e.g., RDI, Radar and LDAR, camera), GPS, wireless (transmitter and receiver) networking components, GSM, RFID, Bluetooth data obtained as historical, present, and predictive traffic. The layer of data collection can be classified as historical.

The ITS framework to efficiently use the collected data to develop improved monitoring strategies. Then this data is analyzed and stored to decide in real-time better route planning, navigation. Another critical concern is how the gathered information will be examined and processed to decide and forecast. To this end, different parametric and nonparametric techniques that are covered in the

Fig. 2. Machine learning-based traffic congestion monitoring.

(5)

following sections are available. In the application layer, the expected effects are used to change the state of transport.

Industrial robots provide more consistency and quality for manufacturers to carry out repetitive tasks. Its behavior is predictable and its motion accurate. This means they can produce products of high quality, with little change and consistency, compared to their people.

Transport plans are necessary to ensure better, more speedy, simple, accessible, economical, and ecological movement for people and goods in the operation, delivery, and management of facilities and services for transport modes. The detection of objects is a computing technology for human-robot interaction and image processing that deals with visual photographs and videos of sem´antic objects of a particular class. The purpose of traffic control equipment is to assure road safety by ensuring that both motorized and unmotorized movement is properly and predictably operated in the road transport system. These sensors instruct drivers when and how to reach them. Navigation is a study area that focuses on controlling and tracking the passage from one location to another craft vehicle. The role of the navigator compared with the identified positions requires all navigation strategies. Congestion management is a way to track the mechanism by which the overall volume of data accessing the network is controlled to maintain an optimal level of traffic. This is done to avoid a congestive breakdown of the telecommunications network.

i) Support vector machine

Support Vector Machine (SVM) specializes in working with small, nonlinear, and high-dimensional samples. The SVM’s fundamental concept is identifying two hyperplanes for proper differentiation of the two data groups while maintaining a classification interval of maximum width. The non-linear transformation φ(y)is utilized to map the nonlinear problem to a linear classification issue in high-dimensional spaces. The support vector regression algorithm (SVR) is based on the basic concept of the SVM to generalize it to the regression issue. SVR overcomes conventional machine learning architectures like neural networks inherent in their deficiencies. It will, in theory, ensure the optimal global and has a transparent model and strong capacity to encourage small samples. support vector regression can solve issues like small quantities, nonlinearity, high measurement, and local optimum conditions and has been used to forecast short-term traffic flows.

Let g be the hyperplane classification and μdenotes normal vectors for a particular nonlinear transformation function. A preview of a training sample set has been shown below:

S={

(Y1,X1),(Y2,X1),…, ( Yj,Xj

)} (1)

H(y) =Z.φ(y) − a=0 (2)

As shown in equation (1), the previous function Shas been determined. Y1 is an input function. X1is an output function. In the above Fig. 3.Traffic flow prediction system.

(6)

formulation Yj∈O^M,Xj∈ {+1,− 1}finds the optimized E(y)function, Xj =E(Yj)daily can be achieved both in the sample of training and in the testing collection, (Yj+1,Yj+2,…,Yn). Then classification Hhyperplane the two samples in a high-dimensional space satisfies the requirements.H(y)is a two samples satisfaction with hyperplane,ais an offset. As deliberated in equation (2), high dimensional space separating hyperplane has been explored. Standardize the vector coefficient φ to satisfy every sample h(y) = 1, and correctly classify every sample.

Figure 4 shows the input variable and optimizing function. SVMs presume the data for which it operates is normally in a range of 0- 1 or -1-1. It is really important to normalize feature vectors until them are supplied to the SVM. Optimization is the problem of finding the number of inputs for an objective function, resulting in an assessment of maximum or minimum functions. It’s the challenge behind several algorithms, from suitable models for logistic regression to machine learning training. Decision function is a tool present in the sklearn machine-learning system classifier{SVC, Logistic Regression}. It tells us how comfortably each Y test value of the classifier is positive or negative (large-magnitude positive) or negative (large-magnitude negative value).

Xj

(Z.φ( Yj

)− a)

≥1 (3)

Equation (3)demonstrated the input variable criteria. Xj is an output variable criteria, Z is a normalized vector coefficient.φ(Yj)is a nonlinear transformation with an input variable, ais an offset. The classification interval is currently equivalent to _‖Z‖²and a high dimension function space convex programming problem is translated into a quadric programming problem of the order of duality as stated below:

maxσ Z(σ) = − 1

2σ^SPσ+e^S (4)

As found in equation (4), a Lagrange multiplier has been obtained. In the formula, σ= (σ1,σ2,…, σm)^εσjis the Lagrange multiplier respective to the inequality constraint Xj(Z.φ(Yj) − a) ≥1Pis aquadratic programming problem, e^Sis an exponential function.

Suppose the training set S= {(Y₁,X₁),(Y₂,X₁),…, (Yj,Xj)},Yj∈O^Cis a d-dimensional input parameter and Xj∈Odenotes the respective output parameter. The support vector regression model can map the training set to high-dimensional feature spacesO^Cvia a nonlinear mapping function φ(y) = {φ(Y₁),φ(Y₂),…,φ(Yj)}. The optimum decision function is defined by

E(y) =Z^Sφ(y) +a,Z∈O^C,C∈O (5)

Equation (5) elaborates on the optimizing function.μ is the weight vector, ais offset, and the fitting functionsE(y)minimizing the systemic risk, the subsequent objective role:

min [1

2Z^sZ+DOempirical

]

(6) As obtained in equation (6), the empirical risk factor with the decision function has been expressed. ¹₂Z^sZ denotes the decision function complexity Oempirical indicates the empirical risk, which signifies training errorsDrepresents the penalty coefficient, which is utilized to equilibrium model training error and complexity. The training errors Oempirical =

(

1 k

)∑_k

j=1

⃒⃒Xj− E(Yj)⃒

⃒can be restrained by τ, and the insensitive loss functions are described by D(Xj,Yj,E(Yj)) =max{0,∑_k

j=1

⃒⃒Xj− E(Yj) − τ^⃒^⃒}

Fig. 4. Input variable and optimizing function.

(7)

ii) Random forest (RF)

The most fundamental component of a random forest is the need to produce several decision trees. K distinct and identically distributed rands of vectors θ₁,θ₂,…,θlneed to generate Ldecision trees.Dis a penalty function and g_lare used to produce the G(Y,θl)to the kth tree where Y is the input vector.

The following is a forest composed of several trees G1(y), G₂(y),…,Gl(y)and two random vectorsY,X. Margin function in equation (7):

Nh(y,X) =AveragelJ(g_l(Y) =X) − max

i∕=XAveragelJ(g_l(Y) =i) (7)

As explored in equation (7) whereNh(y,X)margin function has been performed. g_lis two random series trees. The indicator function is where X is a correct vector classification, and J(.)denotes an indicator function. The edge function symbolizes how the variable Yis classified correctly by the average number of votes in other groups. The higher the edge feature, the greater the trust in the proper classification.

A transportation management system is a logistics platform that utilizes technology to assist businesses in planning, implementing, and optimizing the physical movement of goods both inbound and outbound and ensure that shipments are responsible for compliance.

The Adaboost algorithm is a mathematical theory-based machine learning algorithm. The basic operating theory is that the computer learns several positive and negative samples, such that some main characteristics can be found which differentiate negative and positive samples. Assumed a series of training sample ((Y₁,X₁),(Y₂,X₁),…, (Yj,Xj))Xj = 0denotes negative samples (no car exists), and Xj = 1indicates it is positive samples. Mdenotes the number of training samples; first weight μj =C(j)

For s= 1,…, S;normalized weight Os,j=∑^ZM^s,i

i=1Zs,i for every feature E, train a weak classifier g(y,E,Q,δ):calculate the weighted error rate τe:τe =∑

j

Pj|g(y,E,Q,δ) − Xjthe weak classifier corresponding to every features; then modify the weight consistent with this better weak classifier: Zs+1,j =Zs,jα¹⁻s ^e^j, ej =0 means Yjis misclassified αs =₁₋^τ^s_τ

sThe final strong classifier is in equation (8):

D(y)= {1,∑^S

s=1

σsg(y) ≥1 2

∑^S

s=1

σs

0,

(

σs=log1 αs

)

other

(8)

As described in equation (8) perfect classifier has been expressed. Factors such as the setting and the system for image collection influence image quality, leading to incorrect identification of missing detection. Pre-processing operations on the pictures taken, then images and gamma normalization must be carried out to address too dark or too much light.

Gamma correction can improve or reduce the total image brightness in case of uneven light intensities. Gamma normalization generally takes place utilizing the square root or logarithmic process. Here the pixel value’s square root has been used:

J(y,x) =J(y,x)^δ (9)

As determined, equation (9) pixel value’s square root has been deliberated. In the following J(y,x)denotes pixel values of one point in the graphic, and δindicates the coefficient of transformation 0.5. Various gradient operators have a major influence on detector efficiency. The simplest 1-D, discrete gradient template is a practical approach after experimental research.

For instance, the temperature [-1,0,1] of a Sobel convolution kernel performs well with a 3×3 pixel lower efficiency. The gradient estimation method uses the vertical direction [-1,0,1] and horizontal direction [-1,0,1] as template.

g_y(y,x) =J(Y+1, X) − J(Y− 1,X),g_x(y,x) (10)

As obtained in equation (10), the gradient function has been expressed. The gradient of the gy(y,x),g_x(y,x)points in the horizontal and vertical direction are(y,X). In the image input J(y,x)field is the pixel value of (y,x)point.

Kalman filtering is a bayesian filtering system, and hence Kalman filtering has two essential equations, which corresponds to the following formula:

Ys=EsYs−1+hsθs

Ws=g_sYs+j_sμ_s (11)

Kalman filtering input and weight has been found in equation (11). Es,g_s,Hs,j_s,Psand Osare known vector or matrix. Supposing it is known at the timeS1:

Q(Ys−1W1:S− 1) =M(N(s−1s−1)Q_(s−_1s−₁₎) (12)

The posterior probability has been calculated in equation (12). The connection among the two stages of the state forecasting and update is,

Q(Ys−1W1:S− 1) =M(N(ss−1)Q_(s−_1s−₁₎), Q(YsW1:S) =M(N(ss),Q_(ss)) (13) As shown in equation (13), the prediction state has been described. In general, the Kalman filter approach can provide an optimum

(8)

solution for linear Gaussian versions, and linear operation-dependent calculation speed is very fast. However, the motion of the destination is not linear in practical problems, so the dynamic model is not linear, which does not match Kalman’s filtering procedure.

In this case, it is no longer possible to apply Kalman filtering. An integral operation is performed when computing the posterior probability Q(YsW_1:S)in the Bayesian filtering architecture. This dynamic integration process can be solved, and the Monte Carlo procedure can achieve an optimum solution. The principal aim of Monte Carlo is to assess conditions of an objective function according to this distribution of probability using samples from the distribution of target probabilities:

φ=FQ[E(y)] =

∫

E(y)Q(y)dy (14)

As found in equation (14), target probability distribution has been computed. Where Q(y)is the function for likelihood density FQ is an expectations function, and E(y) represents the function for objective density. Then sample M{Y^j}^M_j=1according to the Q(y)distribution can be sampled, and integral φ can be calculated accordingly:

φ_M= 1 M

∑^M

j=1

E( y^j)

(15) As found in equation (15) normalized vector coefficient function has been obtained. When M→ ∝,φ_M→φerror term order is R(M − 1/2)with φM it can be shown that it does not relate to the size of the random parameterY. The number of samples M can be expanded to reduce the error. A sampling of value is a widely used process. The back probability density of Q(YsW_1:S)can’t be directly achieved in the Bayesian filtering model, so the likelihood density can not be directly dependent on Q(YsW₁:S)samples.

4. Results and discussion

ML-ITMS has been validated based on the complexities of traffic trend characteristics. To determine the brief challenge for a system of traffic lights. The short-term adjustments to traffic conditions can result in large fluctuations of flows in the traffic signal on the lane.

The space relationship between two street segments can shift as well as the traffic light. Besides, one city can adjust the way traffic light is controlled. On the highway, there are various traffic trends in the metropolitan environment. Such two regions have more complex spatial interactions than the standard traffic system. The interaction rate of ML-ITMS is shown in figure. 5.

The metropolitan area has various types of travel on the highway. Such two regions have a more complex cognitive connection than they are in a single transport surrounding. Therefore, it is directed to the prediction system for the road traffic region, and a clear difference must be made among these kinds of traffic trends on the highway. The conversations among these different traffic conditions in the road preview system must be explored by robot interaction. The traffic in the roads can be solved by human-machine interaction, and the prediction rate of the proposed ML-ITMS is shown in figure. 6.

Most advanced prediction technologies concentrate on observable traffic network predictions that help prevent traffic delays and intending traffic routes. The traffic delay is described as the discrepancy among various steps. The traffic delay is less than a few observations at various divergence ranges of the transmitter ranges without a protection function. The overload package and the protection processing delay increase the end-to-end lag due to an additional package overlap. The traffic delay of ML-ITMS is shown in table. 1.

An approach to boost forecast system accuracy is to increase input size, that is to say, to incorporate additional transportation

Fig. 5. The interaction rate of ML-ITMS.

(9)

report forms ML-ITMS often took into account the distance record of various forms of vehicles alongside statistics on weekdays and times of days. The model would be more precise with secret neurons. The developer explored better the impact of data performance levels by the robot interaction. The results revealed that more knowledge is added to the model’s accuracy. However, greater system accuracy makes the proposed system more precise and easy to use. The accuracy of the proposed ML-ITMS is shown in table. 2.

Traffic monitoring systems, normally incorporated with road networks, may provide accurate traffic information. These systems enable vehicles to be detected and ranked in specific locations utilizing sensor data. Due to structural limitations and the growing volume of vehicles, traffic monitoring and control are complicated activities and require specific algorithms and accurate traffic information to be achieved by the human-robot interaction. The details on numbers and types of vehicles help to reduce travel times and emissions. The traffic monitoring rate achieved by ML-ITMS is shown in table. 3.

Reliability influences the number of times passengers to have to wait for a vehicle to arrive at the transit stop and a passenger’s daily consistency in arriving at their destination. Reliability includes both on-time performance and regular progress between suc- cessive vehicles.

For each lane section, the Traffic congestion rate is determined based on their speed limit and volume function, classified by period and day. Implementing robot interaction, the amount of traffic on weekend evenings and the amount of traffic on weekdays decreased;

this primarily affects the related Traffic congestion rate levels explicitly. The traffic congestion rate is reduced using human-robot interaction in road transportation, illustrated in table. 4.

The traffic on the roads depends on the number of vehicles covered by the transportation system. The particular information regarding vehicle usage, the number of automobiles reaching a specific stage, and the average speed reduce traffic congestion. ITMS can determine day and night range traffic on weekdays, and weekend traffic congestion rate

The proposed method achieves the highest traffic monitoring rate when compared to other existing Task-Adaptation in Physical Human-Robot Interaction (TA-HRI), Gesture-based human-robot interaction (GB-HRI), Emotional Processes in Human-Robot Inter- action (EP-HCI).

5. Conclusion

This paper presents ML-ITMS to address various problems and to enhance the safety measures of transport and reliability. To ensure Fig. 6.The prediction rate of ML-ITMS.

Table 1

The traffic delay of ML-ITMS.

Number of Vehicles EP-HCI GB-HRI TA-HRI ML-ITMS

10 56.3 51.3 65.2 41.2

20 54.1 52.6 66.2 42.3

30 51.2 53.8 67.1 43.6

40 58.2 54.7 68.2 46.5

50 59.8 55.6 65.3 47.2

60 62.3 52.3 64.2 45.3

70 62.1 51.4 64.3 44.2

80 61.8 50.2 65.2 43.6

90 63.2 50.8 62.3 41.2

100 62.1 55.2 63.6 41.1

(10)

the traffic condition and other important regulations regarding road transportation systems, ITS was implemented. ITS allows fixing the potential route situation in advance. The main problems with ITS are important to achieve a precise and efficient traffic flow prediction method. The proposed ML-ITMS used a mathematical equation to optimize traffic flow and non-parameter process accuracy estimations. ML approach is the best existing technique of nonparametric method, and it needs less knowledge about the links between different traffic patterns and better viability of nonlinear traffic data. The HCI allowed both consumers and suppliers at both ends of transport systems to address important problems simultaneously. The experimental results of ML-ITMS showed that the traffic monitoring rate could be improved to the rate of 98.6% and improved the traffic flow prediction systems than other current methods.

References

On the design of a human-robot interaction strategy for commercial vehicle driving based on human cognitive parameters.

An effective and powerful traffic forecasting system could provide continual and exact details on-road position focused on past traffic situations as an essential part of ITS.

Desideri, L., Ottaviani, C., Malavasi, M., di Marzio, R., & &Bonifacci, P. (2019). Emotional processes in human-robot interaction during brief cognitive testing.

Computers in Human Behavior, 90, 331–342.

Elhoseny, M., Shehab, A., & Yuan, X. (2017). Optimizing robot path in dynamic environments using genetic algorithm and Bezier curve. Journal of Intelligent & Fuzzy Systems, 33(4), 2305–2316.

Fu, H., Wang, Y., Tang, X., Zheng, N., & Geroliminis, N. (2020). Empirical analysis of large-scale multimodal traffic with multi-sensor data. Transportation Research Part C: Emerging Technologies, 118, Article 102725.

Gao, J., Wang, H., & Shen, H. (2020, August). Machine learning based workload prediction in cloud computing. In 2020 29th international conference on computer communications and networks (ICCCN) (pp. 1–9). IEEE.

Table 2

The accuracy of ML-ITMS.

10 45.6 56.7 70.3 90.2

20 44.3 50.9 75.4 98.90

30 47.8 54.3 78.9 94.5

40 48.9 52.4 79.3 97.8

50 42.1 55.7 75.5 96.5

60 40.3 58.7 73.2 93.4

70 46.8 59.8 74.7 92.1

80 48.9 53.4 77.2 91.3

90 49.2 57.8 70.2 90.3

100 41.2 59.1 73.4 95.6

Table 3

The traffic monitoring rate achieved by ML-ITMS.

10 76.2 65.2 51.3 91.3

20 76.3 66.3 52.6 95.6

30 77.4 67.1 53.8 94.5

40 77.8 68.3 54.7 90.6

50 79.2 69.2 55.6 93.5

60 78.2 70.3 52.3 92.8

70 78.8 72.3 51.4 93.7

80 79.2 76.8 50.2 97.1

90 79.4 77.9 50.8 94.2

100 79.9 79.2 51.3 98.6

Table 4

The traffic congestion rate achieved by ML-ITMS.

10 41.2 51.3 72.3 32.1

20 42.3 52.6 75.2 30.2

30 43.6 53.8 71.3 35.1

40 46.5 54.7 69.2 33.6

50 47.2 55.6 66.2 34.1

60 45.3 52.3 62.11 39.2

70 44.2 51.4 90.3 33.5

80 43.6 50.2 62.1 35.1

90 41.2 50.8 60.1 33.9

100 41.1 51.3 72.3 32.1

(11)

Gao, J., Wang, H., & Shen, H. (2020). Task failure prediction in cloud data centers using deep learning. IEEE Transactions on Services Computing, 1111–1116. https://

doi.org/10.1109/BigData47090.2019.9006011

... & Guo, X., Aviles, G., Liu, Y., Tian, R., Unger, B. A., Lin, Y. H. T., & Kampmann, M. (2020). Mitochondrial stress is relayed to the cytosol by an OMA1–DELE1–HRI

pathway. Nature, 579(7799), 427–432.

Jindal, A., Aujla, G. S., & Kumar, N. (2019). SURVIVOR: A blockchain-based edge-as-a-service framework for secure energy trading in SDN-enabled vehicle-to-grid environment. Computer Networks, 153, 36–48.

Khoramshahi, M., & Billard, A. (2019). A dynamical system approach to task adaptation in physical human-robot interaction. Autonomous Robots, 43(4), 927–946.

Lin, K., Li, Y., Sun, J., Zhou, D., & Zhang, Q. (2020). Multi-sensor fusion for a body sensor network in a medical human-robot interaction scenario. Information Fusion, 57, 15–26.

Manogaran, G., Baskar, S., Hsu, C. H., Kadry, S. N., Sundarasekar, R., Kumar, P. M., & Muthu, B. A. (2021). FDM: Fuzzy-optimized Data Management Technique for Improving Big Data Analytics. IEEE Transactions on Fuzzy Systems, 29(1), 177–185.

Manogaran, G., Shakeel, P. M., Priyan, R. V., Chilamkurti, N., & Srivastava, A. (2019). Ant colony optimization-induced route optimization for enhancing the driving range of electric vehicles. International Journal of Communication Systems, e3964. https://doi.org/10.1002/dac.3964

Nasimi, M., Habibi, M. A., & Schotten, H. D. (2020). arXiv preprint. Platoon–assisted vehicular cloud in VANET: vision and challenges. arXiv preprint arXiv:2008.10928.

Neto, P., Sim˜ao, M., Mendes, N., & &Safeea, M. (2019). Gesture-based human-robot interaction for human assistance in manufacturing. The International Journal of Advanced Manufacturing Technology, 101(1), 119–135.

Nguyen, N. T., Liu, B. H., Pham, V. T., & Huang, C. Y. (2016). Network under limited mobile devices: A new technique for mobile charging scheduling with multiple sinks. IEEE Systems Journal, 12(3), 2186–2196.

Preeth, S. S. L., Dhanalakshmi, R., & Shakeel, P. M. (2020). An intelligent approach for energy efficient trajectory design for mobile sink based IoT supported wireless sensor networks. Peer-to-Peer Networking and Applications, 13(6), 2011–2022.

Priyan, M. K., & Devi, G. U. (2018). Energy-efficient node selection algorithm based on node performance index and random waypoint mobility model on the Internet of vehicles. Cluster Computing, 21(1), 213–227.

Ramprasad, L., & Amudha, G. (2014, February). Spammer detection and tagging based user generated video search system—A survey. In International Conference on Information Communication and Embedded Systems (ICICES2014) (pp. 1–5). IEEE.

Shahriar, M. R., Al Sunny, S. N., Liu, X., Leu, M. C., Hu, L., & Nguyen, N. T (2018, June). MTComm based virtualization and integration of physical machine operations with digital-twins in cyber-physical manufacturing cloud. In 2018 5th IEEE International Conference on Cyber Security and Cloud Computing (CSCloud)/2018 4th IEEE International Conference on Edge Computing and Scalable Cloud (EdgeCom) (pp. 46–51). IEEE.

Shakeel, P. M., Arunkumar, N., & Abdulhay, E. (2018). Automated multimodal background detection and shadow removal process using robust principal fuzzy gradient partial equation methods in intelligent transportation systems. International Journal of Heavy Vehicle Systems, 25(3-4), 271–285.

Shakeel, P. M., Baskar, S., & Selvakumar, S. W. P. C. (2019). Retrieving multiple patient information by using the virtual MIMO and path beacon in wireless body area network. Wireless Personal Communications, 108(4), 2359–2370. https://doi.org/10.1007/s11277-019-06525-5

Sheikh, M. S., Liang, J., & Wang, W. (2020). Security and privacy in vehicular ad hoc network and vehicle cloud computing: a survey. Wireless Communications and Mobile Computing, 2020, Article 5129620. https://doi.org/10.1155/2020/5129620

Sirohi, D., Kumar, N., & Rana, P. S. (2020). Convolutional neural networks for 5G-enabled intelligent transportation system: A systematic review. Computer Communications, 153, 459–498.

Sivaparthipan, C. B., Muthu, B. A., Manogaran, G., Maram, B., Sundarasekar, R., Krishnamoorthy, S., & Chandran, K. (2019). An innovative and efficient method of robotics for helping the Parkinson’s disease patient using IoT in big data analytics. Transactions on Emerging Telecommunications Technologies, e3838. https://doi.

org/10.1002/ett.3838

Ullrich, D., Butz, A., & Diefenbach, S. (2021). The Development of Overtrust: An Empirical Simulation and Psychological Analysis in the Context of Human-Robot Interaction. Frontiers in Robotics and AI, 8, 44.

Vasconez, J. P., Kantor, G. A., & Cheein, F. A. A. (2019). Human-robot interaction in agriculture: A survey and current challenges. Biosystems Engineering, 179, 35–48.